OpenAI

Recently shipped GPT-6 Astra, a new top-tier API model, alongside Ultrafast mode, healthcare data connections, and expanded ChatGPT Ads.

GPT-6 Astra is the new flagship for complex end-to-end work. The Responses API model handles reasoning, coding, computer use, research, and document creation. It drops support for none reasoning effort, custom temperature/top_p/logprobs, and requires the Responses API for tool calling. New controls for long-running work include async tool calling, mid-turn steering over WebSockets, and changing reasoning effort mid-conversation. Misalignment monitoring asynchronously checks agent work.

API platform gained security and infrastructure features. Mutual TLS and X.509 workload identity federation are generally available, configurable in the Platform console. Connections to api.openai.com now support IPv6. Regional processing can be selected per request via a prefixed domain, and API errors now distinguish traffic spikes (slow_down 429s) from overload (503s).

GPT-5.6 Sol pricing dropped 20% for input and 33% for output tokens, now $4/$20 per million through at least November 21.

ChatGPT connected to healthcare data for clinicians, and event-triggered scheduled tasks launched from Gmail, Slack, and GitHub events. ChatGPT Ads hit a $1B annualized run rate.

Transcription models deprecated. whisper-1, gpt-4o-transcribe, and related models shut down February 26, 2027 in favor of gpt-live-transcribe/gpt-transcribe; the Assistants API shut down August 26.

Codex and SDK updates: Codex CLI 0.153.0 added Vim undo/redo, plugin marketplaces, and GPT-6 Astra support; codex mcp-server is deprecated. Python SDK v3.0.0 defaults to HTTPX2 (breaking), and all SDKs added GPT-6 Astra features.